GENEALOGY-DNA-L Archives

Archiver > GENEALOGY-DNA > 2010-03 > 1269454632


From: David Faux <>
Subject: Re: [DNA] Comparing Ancestry Painting (23andme) to AncestralOrigins Genome B...
Date: Wed, 24 Mar 2010 11:18:58 -0700
References: <2d80.5f9729fa.38db36e0@aol.com>
In-Reply-To: <2d80.5f9729fa.38db36e0@aol.com>


Ann,

I am as much interested in the biogeographical ancestry component as the
percentage sharing with cousins that one can obtain by these various tests.
I have one Six Nations ancestor, Catharine Brant, married to Lt. John Young,
born 1747, for whom I have an abundance of genealogical information (as I do
for his second wife Margaret Hill Kayadontyi born 1749). The First Nations
paper trail continues to 1896 when a first cousin of one of my ancestors
died and his obituary gave precise details of his chiefship and installation
procedure (an embarassment of riches).

Although discredited, the DNAPrint Genomics test was the only one to use
AIMS (ancestral informative markers). Here, knowing some of the problems
with this test (false positives for Pakistanis and Greeks for example), I
tested four family members. The results ranged from 8% Native American for
the descendant with the closest connection to Catharine (great great great
grandson), down to 2% for myself at 1/256. The sequence was correct if not
the exact percentages.

One problem is that on both sides of Catharine's family she is highly
admixed. Her paternal grandparents resided in a two story house and
described to be living a "white middle class" lifestyle. Her maternal side
included an uncle named "White Hans", "a whitish Indian", and a first cousin
John "Blue Eyed" Green Aronghyengtha. Hence culturally a Mohawk but
biologically an admixture with perhaps more than 50% European. This means
that the results are likely to be dodgy. However this did not affect the
inheritance pattern in the Owachira (maternal lineage) whereby when Francis
Cotter Tahatonne and family left in 1817 to join the Wyandot at Anderdon and
Sandusky, and my ancestor's first cousin took over that Mohawk principle
title in that same year although only 17 years old.

Unfortunately the results of testing members of two of the four children of
Catharine has produced results from 23andMe that do not mesh with the
genealogy - at least not at all well. The results from Dr. McDonald's
testing are, however, consistent with the genealogy with again the same
individual having the highest Amerind percentage. What is important here is
that Dr. McDonald uses BEAGLE to analyze the data. This means that instead
of relying on the Han Chinese as a "proxy" to Native American he uses actual
NA samples from the CEPH-HGDP panel.

Unfortunately the deCODEme results can be interpreted in many ways (I
suppose to fit one's preconceived notions if one wished) due to their odd
display with percentages from all three groups on each segment. As you say,
this makes little to no sense. It only makes sense when one can see
visually a distinct block of Asian or African - but this is a subjective
call, which should not be the case. I have tried to find my wife's Asian
block in deCODEme without success but did find another with the latter
presentation. I am losing faith in all of these measures and find that I
will fall back on genealogy as the arbitrator.

None of these groups fully address the X chromosome, although deCODEme does
provide some output. The "gross" percentage needs to be ignored and only
those blocks which conform to what would be expected based on genetic
principles explored. This means a block reaching half way across the X for
females and all the way across for males with say African or Asian.
Although I have both on my X, I am well aware that this could still be
essentially noise and await a more sophisticaled investigation, as was done
for me by Anders Palsen with PLINK.

A large potenial problem is that using the Han Chinese as a proxy for Native
American may tend to give results that are difficult to interpret. I have
no doubt that for some this works fine. In my sample of registered Native
American (Tribal members), the percentages are 5% to 20% "Asian". This is
likely on the low side based on the genealogy. I seriously don't know what
I can trust and what I can ignore. It seems almost impossible to
differentiate signal from noise.

Most African Americans has some (e.g., 10%) Asian on 23andMe. It would be
easy to say that it is undocumented Native American. However the 23andMe
scientists have found that to their amazement, some Africans (particularly
those away from the Nigerian area) have a percentage of Asian showing in the
"Ancestry Paintings". So how is an African - American to know what is real
NA and what is an artifact (but perhaps useful in more closely pinpointing
an African homeland).

The fact is that 23andMe has not updated their Ancestry Painting since 23
April 2008 and I have a very thick binder or relevant articles which suggest
that a great deal more accurate information could be harvested by keeping
current. Also after two years of promises the X is still not painted. This
is of particular interest to me since I am one of the few in my extended
family from Catharine Brant who is in the X - line, and only 4 meioses from
her. At the moment I do have the analysis of Anders Palsen and the diagram
on deCODEme to suggest what might be observed if 23andMe paints the X - but
I hope for multiple sources of evidence that are, hopefully, consistent.

It is rather a shame that FTDNA don't explore biogeographical ancestry with
their new product. I am not looking for more cousins (I have plenty enough
thank you), but rather want to use surgical precision in chosing cousins
with known genealogies to compare and contast the two lines of primary
evidence of genealogists, paper trail genealogy and genetics. Right now I
am not much further ahead than I was in the days of DNAPrint. Pity - and
very discouraging - especially were I to tally the money I have spent on
these tests for family members over the years balanced against what I
obtained for my investment.

On the up side I am jazzed at what 23andMe has been about to say about the
amount of sharing between the various cousins in the study compared with
predicted percentages. It has also helped to clear up a genealogical
mystery about one branch being descendants of a half sibling to my line.
The numbers help immeasurably, assuming that the sample size is reasonably
large - we have had a few anomalies which to this day I still cannot explain
- and "cousins" out there with more sharing than my known cousins, but
nothing in the paper trail to explain it. Chaos rules supreme.

David K. Faux.

On Wed, Mar 24, 2010 at 2:35 AM, <> wrote:

> FTDNA doesn't analyze biogeographic ancestry.
>
> Lots of people on DNA-Forums have posted comparisons of 23andMe and
> deCODEme. Virtually everyone who is reported as 100% European at 23andMe
> receives a
> small percentage of Asian and African at deCODEme. This is due to the
> different approaches taken for analyzing the same set of HapMap data. Both
> companies analyze the chromosomes segment by segment. This is the big
> innovation
> compared to DNA Print, which was based on a small number of Ancestry
> Informative Markers scattered over the genome.
>
> The main difficulty is that there are very few markers that are 100%
> informative (found in everyone in one population and no one in other
> populations).
> For each segment, deCODEme will assign a probability to the origin, say a
> 7% probability that the segment is of Asian origin and 4% of African
> origin.
> 23andMe reasons that a small segment is unlikely to be composed of three
> different origins, so it picks the one or two most likely origins (since
> you
> have two parents).
>
> 23andMe results seem to be sensible for people with documented admixture
>
> http://spittoon.23andme.com/2008/06/13/a-beautiful-ancestry-painting/
>
> or for people where admixture is plausible based on the population history
> (e.g. Puerto Ricans tend to show all three components). For most cases I've
> seen, if the 23andMe painting does show admixture, deCODEme will also
> recognize the same segments. However, it seems to me that the deCODEme
> percentages
> for 23andMe's 100% Europeans are so similar for everyone across the board
> that they are hardly informative about an individual's ancestry.
>
> One last point -- deCODEme is apparently doing Monte Carlo simulations, so
> some people who uploaded their 23andMe results multiple times got different
> answers on each run.
>
> Ann Turner
> Disclosure: I have a consulting arrangement with 23andMe, but the
> statements expressed above are my own and not official policy of 23andMe.
>
>


This thread: